CN115943216A - Integrated dielectrophoresis-capture and nano-trap transfer method for realizing double-poisson single-cell RNA sequencing - Google Patents

Abstract

The present invention provides systems and methods for single cell RNA sequencing. An embodiment of the method of the invention comprises the steps of: aligning a micro-well array over a Dielectrophoresis (DEP) single-cell trapping nano-well array; loading a plurality of cells into a nanowell; applying electricity to the nanowell array to trap a number of cells equal to the number of electrode pairs in at least one nanowell of the nanowell array; interrupting power to the array of nanowells to transfer the loaded cells from the nanowells to the microwells; loading a plurality of barcoded beads into a microwell such that a single bead occupies each cell-loaded microwell; capturing RNA from the cells and retrieving the RNA-loaded beads; and sequencing the captured RNA.

Description

Integrated dielectrophoresis-capture and nano-trap transfer method for realizing double-poisson single-cell RNA sequencing

Cross Reference to Related Applications

The present application claims priority from U.S. provisional patent application No. 63/027,582, filed on 20/5/2020, the contents of which are incorporated herein by reference in their entirety.

Technical Field

Current high throughput single cell RNA sequencing (scra-seq) technology is based on random pairing of cells and barcoded beads in nanoliter droplets or wells. It is limited by the mathematical principle of poisson statistics such that the utilization of either cells or beads or both does not exceed about 33%. Although multifunctional designs of microfluidics or microtrap for high throughput loading of beads exceed the poisson limit, subsequent encapsulation of single cells is still determined by random pairing, which represents a fundamental limitation in the field of single cell sequencing.

Disclosure of Invention

In certain aspects, the invention provides methods for single cell RNA sequencing, the methods comprising the steps of: aligning a micro-well array over a Dielectrophoresis (DEP) single-cell trapping nano-well array; loading a plurality of cells into a nanowell; applying electricity to the nanowell array to trap a number of cells equal to the number of electrode pairs (quanta) in at least one nanowell of the nanowell array; interrupting power to the array of nanowells to transfer the loaded cells from the nanowells to the microwells; loading a plurality of barcoded beads into a microwell such that a single bead occupies each cell-loaded microwell; capturing RNA from the cells and retrieving the RNA-loaded beads; and sequencing the captured RNA.

In some embodiments, the micro-well array comprises wells having a diameter of 50 μm. In some embodiments, the diameter of the nanowell is selected from: 10 μm, 15 μm and 20 μm. In some embodiments, the RNA is sequenced using one or more techniques, including PCR. In some embodiments, the cells are loaded into the nanowell by applying a first alternating potential.

In some embodiments, the method further comprises loading a plurality of second cell types into the nanowell. In some embodiments, the second cell type is loaded using a second alternating potential.

In some embodiments, the method further comprises inverting the aligned array such that the array of microtells is located below the array of nanowells.

In certain aspects, the invention relates to a DEP trapping nanowell transfer (dTNT) system comprising: a single cell trapping nanowell array and a microwell array pre-aligned over the nanowell array, wherein the microwell array is aligned using a micro-aligner device.

In some embodiments, the single-cell trapping nanowell array comprises wells having diameters selected from 10 μm, 15 μm, and 20 μm.

In some embodiments, the micro-aligner apparatus is adapted and configured to align wells of the nano-well array with wells of the micro-well array.

Drawings

For a fuller understanding of the nature and desired objects of the present invention, reference should be made to the following detailed description taken together with the accompanying figures wherein like reference numerals represent corresponding parts throughout the several views.

FIG. 1 depicts an exemplary Design of (DEP) -trapping nano-trap transfer (dTNT) sequencing (dTNT-seq) contemplated herein. It shows a schematic of the DEP-based single cell mRNA sequencing platform and workflow. The two separate layers are pre-aligned and assembled using custom designed manipulators. After cell loading, the single cell was actively trapped by positive DEP into the nanowell. The entire device is then inverted and the cells are transferred by gravity into a larger microwell, followed by barcoded bead loading, cell lysis, and mRNA capture by DNA oligomers on the bead surface, each bead containing a cell barcode and a Unique Molecular Identifier (UMI). The beads were collected and pooled, and the mRNA was reverse transcribed in large amounts to form single cell transcriptomes (STAMPs) attached to the microparticles. Amplification and tagging (tagmentation) was then performed to prepare sequencing libraries. Sequencing data for transcriptome alignments were performed to generate gene expression matrices for downstream data analysis.

Fig. 2A-2E depict exemplary designs and configurations of dTNT devices as contemplated herein. Fig. 2A depicts a top view of an array of DEP nanowells for single cell trapping. Fig. 2B depicts an exemplary microscope image of the fabricated electroactive DEP array. Fig. 2C depicts an exemplary 3D optical surface profiler image of a cell capture nanowell. Fig. 2D depicts a cross-sectional view of an assembled dTNT device, including a larger microwell layer on top, DEP array chip on bottom, and PDMS gasket in between to form flow channels for loading cells, beads, and all reagents. Figure 2E depicts a photograph of a pre-aligned two-layer dTNT device assembled using a home-made aligner. Right inset: optical image, showing an enlarged view of a representative region of the dTNT device.

Figures 3A-3E illustrate the evaluation of single cell capture, occupancy, transfer efficiency and bead loading based on DEP. Fig. 3A depicts a statistical analysis of the number of cells in 3080 nanowells in 35 regions imaged by fluorescence. Overall, more than 90% of the nanowells were occupied by single cells and the double rate (doublt rate) was less than 2%. Fig. 3B depicts fluorescence images of single cells (green) captured using a 10 μm deep nanowell. Figure 3C depicts cell capture performance as a function of nanowell depth. In the current study, the effects of 5, 10, 15 and 20 μm depths were studied. The 10 μm nanowell produces optimal single cell trapping with negligible double rate. Fig. 3D depicts a fluorescence image of a single cell transferred into a large microtrap. After inverting the dTNT device, an average of 82% of the captured cells were successfully transferred. Figure 3E depicts barcoded beads loaded into a microwell at a single bead occupancy of approximately 100% due to the exclusion of microwell size and the ability to move the beads back and forth in the flow channel.

FIGS. 4A-4E depict single cell RNA sequencing of a mixed sample of species using dTNT-seq. Fig. 4A depicts fluorescence images of mouse 3T3 (green) and human HEK (red) cells on DEP nanowell arrays. Figure 4B depicts sequencing reads mapped to human and mouse genomes. (FIGS. 4C and 3D) Violin plots of # s showing genes or transcripts detected in single cells (center line: median; limits: first and third quartiles; whiskers, + -1.5 IQR; dots, > 1.5 IQR) FIG. 4E depicts the efficiency of human gene capture compared to gene capture in Seq-Well using PBMC samples.

Fig. 5A-5D depict a graph-based unsupervised cluster analysis and comparison with a non-electrode method (scFTD-seq). Fig. 5A depicts UMAP visualization of two major species-specific groups generated using dTNT, where each group has three different single-cell clusters. FIG. 5B depicts the heat map expression of the first 5 differentially expressed gene markers in each cluster in a dNT-seq. FIG. 5C depicts UMAP visualization of single-cell transcriptomes generated using scFTD-seq. As with dTNT, two major species-specific groups were identified, with two larger subpopulations in each group. FIG. 5D depicts the number of cells per major cluster in dNT-seq and scFTD-seq.

FIGS. 6A-6D depict a comparison of the biological processes of clusters identified between dNT-seq and scFTD-seq by GSEA analysis. The top 10 GO items enriched in clusters (fig. 6A) DEP _ human 0; (fig. 6B) human 0; (fig. 6C) DEP _ mouse 1; (FIG. 6D) mouse 1. Representative GSEA enrichment maps and the distribution of marker genes defining each cluster are also shown. The GO term is ordered by Normalized Enrichment Score (NES) generated by GSEA.

Fig. 7 depicts a complete fluorescence image after transfer of the captured monocytes. By inverting the device, an average of 82% of the captured single cells were successfully transferred to the larger microwell below.

Figure 8 depicts a complete fluorescence image of barcoded bead loading for mRNA capture. Due to the exclusion of microwell size and the ability to move beads back and forth, barcoded bead loading rates can approach 100%.

FIG. 9 depicts an exemplary workflow of dTNT-seq operation and processing time for each step.

Figure 10 depicts large area fluorescence images captured by species-mixed human and mouse cells of DEP. Here, mouse (3T 3) cells were labeled with green fluorescent dye and Human (HEK) cells were labeled with red fluorescent dye.

FIGS. 11A-11C depict the evaluation of single cell resolution and transcriptome quality. Figure 11A shows that for most cells, more than 90% of the transcripts are aligned with the species-specific genome. FIG. 11B shows the relationship between the number of transcripts and the percentage of mitochondrial genes. FIG. 11C shows the relationship between the number of transcripts and the number of genes. Cells were filtered according to gene count (200 to 5000) and percentage of mitochondrial genes (less than 10%) to exclude low quality cells or potential cell doublets.

FIGS. 12A-12D illustrate the qualification and linear dimensionality reduction (PCA) of highly variable features. Figure 12A depicts 2000 genes in the data set that showed high intercellular variation (i.e., they are highly expressed in some cells and low expressed in other cells) were selected for downstream analysis and the top 10 most highly variable genes were labeled. Fig. 12B depicts cells visually defining PCA. FIG. 12C is a Jackstraw diagram. In the JackStraw plot, significance dropped dramatically after the first 7 PCs. FIG. 12D shows that in the Elbow plot, a "bend" near PC8-9 can be observed. It is recommended to be higher when this parameter is chosen according to the saurat protocol, so we choose 10 PCs as input for performing UMAP clustering.

Fig. 13A and 13B depict exemplary heat maps of the first 15 enriched genes found to define each cluster of (fig. 13A) human species and (fig. 13B) mouse species.

14A-14C depict the first 10 GO terms (FIG. 14A) DEP _ human 1 enriched in clusters; (fig. 14B) human 1; (FIG. 14C) DEP _ human 2. GSEA enrichment maps for the first 3 GO entries are also shown. The GO terms are ordered by Normalized Enrichment Scores (NES) generated by GSEA.

15A-15C depict the top 10 GO terms (fig. 15A) DEP _ mouse 0 enriched in clusters; (FIG. 15B) mouse 0; (FIG. 15C) DEP _ mouse 2. GSEA enrichment maps for the first 3 GO entries are also shown. The GO terms are ordered by Normalized Enrichment Scores (NES) generated by GSEA.

Figure 16 shows the design patterns and workflow of two types of cell pairing for studying cell-cell interactions by scRNA-seq.

Fig. 17A and 17B demonstrate an exemplary "rooftop DEP nanowell array" embodiment of the invention, in which the nanowell array directly captures the top single cell without inversion. Fig. 17A depicts a schematic of this method. Figure 17B depicts a cross-sectional illustration of a top-loaded nanowell array with cells.

Fig. 18 depicts an exemplary embodiment in which an addressable "rooftop DEP nanowell array" device is designed to enable flexible manipulation of target cells.

Definition of

The invention may be best understood by reference to the following definitions.

As used herein, the singular forms "a", "an" and "the" include plural references unless the context clearly dictates otherwise.

Unless specifically stated or apparent from the context, as used herein, the term "about" is understood to be within the normal tolerance of the art, e.g., within 2 standard deviations of the mean. "about" can be understood to be within 10%, 9%, 8%, 7%, 6%, 5%, 4%, 3%, 2%, 1%, 0.5%, 0.1%, 0.05%, or 0.01% of the stated value. All numerical values provided herein are modified by the term "about," unless otherwise apparent from the context.

As used in the specification and claims, the terms "comprising", "containing", "having", and the like may have the meaning given to them by U.S. patent law, and may mean "including" and the like.

The term "or" as used herein is to be understood as being inclusive unless specifically stated or apparent from the context.

Ranges provided herein are to be understood as shorthand for all values falling within the range. For example, a range of 1 to 50 is understood to include any number, combination of numbers, or sub-range selected from 1,2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, or 50 (and fractions thereof, unless the context clearly dictates otherwise).

Detailed Description

The present invention provides integrated Dielectrophoresis (DEP) -trapping nanowell transfer (dTNT) systems and methods for high-throughput single-cell RNA sequencing (scRNA-seq). In particular, the present invention provides an integrated Dielectrophoresis (DEP) -capture nanowell transfer (dTNT) method, referred to as dTNT-seq, whereby cell capture and bead loading are performed in a poisson manner to facilitate scRNA-seq.

dTNT sequence system

Referring now to FIG. 1, the dTNT-seq system of the present invention includes a DEP-based single cell mRNA sequencing platform. The system includes a micro-well array slide, a DEP nano-well array slide, and a micro-aligner for precisely aligning wells of the nano-well array and the micro-well array slides when assembled.

The microwell array slide includes a well array sized to accommodate DNA barcoded beads used in scRNA-seq analysis. Each microwell in the array may be sized to have a diameter of about 50 μm. In some embodiments, the diameter of the microwells ranges from about 20 μm to about 30 μm, about 30 μm to about 40 μm, about 40 μm to about 50 μm, about 50 μm to about 60 μm, about 60 μm to about 70 μm, about 70 μm to about 80 μm, and any and all increments therein. In certain embodiments, each microtrap in the array has a depth of about 50 μm. The depth of the micro-wells may be about 20 μm to about 40 μm, about 40 μm to about 60 μm, about 60 μm to about 80 μm, and any and all increments therebetween. Embodiments of the micro-wells have a pitch of about 100 μm. The pitch of the micro-wells may be about 60 μm to about 80 μm, about 80 μm to about 100 μm, about 100 μm to about 120 μm pitch, and any and all increments therebetween.

The micro-well array slide may be made of any suitable material as understood in the art. For example, embodiments of the micro-wells are made of Polymethylmethacrylate (PMMA). In some embodiments, the microwells are fabricated directly in the SU-8 layer coated on the PMMA substrate. In some embodiments, silicon or glass is used as the substrate material for fabricating the SU-8 microwells.

The micro-wells are positioned along a series of microfluidic channels positioned on the substrate. The substrate may also include fluid access holes through the silicon or glass for introducing the beads.

The nanowell array slide included a DEP trap. Embodiments of a nanowell array slide include a nanowell array sized to isolate a single cell. The diameter of each well is about 20 μm. Embodiments of the nanowell have a diameter of up to about 5 μm, from about 5 μm to about 10 μm, from about 10 μm to about 15 μm, from about 15 μm to about 20 μm, from about 20 μm to about 25 μm, and any and all increments therein. In some embodiments, the diameter is 5 μm, 10 μm, 15 μm, or 20 μm. The depth of the nanowell is about 20 μm. Embodiments of the nanowell may have a depth of about 5 μm to about 10 μm, about 10 μm to about 15 μm, about 15 μm to about 20 μm, about 20 μm to about 25 μm, and any and all increments therebetween. The spacing of the nanowells matches the spacing of the microwells on the microwell array slide. For example, the nanowells may have a pitch of about 100 μm. The nanowells are aligned along microchannels formed in the slide.

The micro-well array slide and the DEP nano-well array slide can be precisely aligned and assembled. In some embodiments, when assembled, the gasket is positioned between two slides to form a flow channel for loading cells, beads, reagents, and the like. The thickness of the gasket may be about 100 μm. The thickness can be about 50 μm to about 75 μm, about 75 μm to about 100 μm, about 100 μm to about 125 μm, about 125 μm to about 150 μm, and any and all increments therein. The gasket may be constructed of any suitable material understood in the art, including, for example, PDMS.

Each of the nanowell array slide and the microwell array slide may include a plurality of wells. For example, the array slide may have up to 2000 wells, about 2000 wells to about 2200 wells, about 2200 wells to about 2400 wells, about 2400 wells to about 2500 wells, about 2600 wells to about 2800 wells, about 2800 wells to about 3000 wells, about 3000 wells to about 3200 wells, about 3200 wells to about 3400 wells, about 3400 wells to about 3600 wells, about 3600 wells to about 3800 wells, about 3800 wells to about 4000 wells, about 4000 wells to about 4200 wells, about 4400 wells to about 4400 wells, about 4400 wells to about 4600 wells, about 4600 wells to about 4800 wells, about 4800 wells to about 5000 wells, and any and all increments therein.

The micro-and nano-well array slides may be assembled such that the wells in each slide are perfectly aligned. In some embodiments, the two slides are aligned using the aligner shown in fig. 2E. The aligner contemplated herein ensures alignment of the nanowell and the microwell, which ensures efficient cell trapping and transfer.

In certain embodiments, the system of the present invention is assembled such that the nanowell array slide is at the bottom of the microwell array. In this case, once the cells were loaded into the DEP nanowell array slide, the assembled system was inverted. The microwell array slide and the nanowell array slide may also be aligned by one or more pins, slots, bosses, or other mechanical means. The microwell array slide and the nanowell array slide may also be aligned using imaging (e.g., optical imaging) prior to being affixed to each other (e.g., with an adhesive).

In some embodiments, the system of the present invention is assembled such that the nanowell array slide is on top of the microwell array slide. In this case, the cells are loaded into the nanowell array while in a top "roof" position or orientation.

Method

The present invention provides methods for high throughput single cell sequencing using the dTNT device contemplated herein.

Referring now to fig. 1 and 9, embodiments of the method are summarized in a schematic and workflow diagram of the dTNT-seq of the present invention.

Embodiments of the method include first precisely aligning an array of micro-wells over an array of DEP nano-wells. The wells of each array slide are aligned using a micro-aligner device. In some embodiments, a gasket is positioned between the micro-well array and the nano-well array to form a channel between the two slides.

Embodiments of the method include loading a loaded cell into a microfluidic channel positioned along a nanowell slide. The cells may include any cell type understood in the art, as preferably used. For example, the cells may include primary cells, immortalized cells, stem cells, and the like. Cells may include cells isolated from any suitable species known in the art, including, for example, murine (murine), human, murine (ratus), rabbit, bovine, porcine, canine, equine, and the like. Cells may include cells isolated from one or more suitable tissues including, for example, vascular tissue, blood, muscle tissue, neural tissue, bone tissue, respiratory tissue, prostate tissue, heart tissue, pancreatic tissue, and the like, including normal tissue, cancerous tissue, and/or cells. The cells may comprise one or more cell types including mouse embryonic fibroblast (NIH 3T 3), human embryonic kidney (HEK 293) cells, human Peripheral Blood Mononuclear Cells (PBMCs), including human monocyte-like cells (U937), lung cancer cells (NCIH 1975), prostate cancer cells (DU 145 and PC 3), breast cancer cells (MCF-7), and HeLa cells, and one or more combinations thereof.

Embodiments of the method include applying a voltage to an array of DEP nanowells. The applied voltage may comprise an applied alternating potential. Embodiments of the applied alternating potential may include a peak-to-peak (Vp-p) potential of 4V, where V is the peak-to-peak voltage. The electrical potential can include up to 4V, about 4V to about 6V, about 6V to about 8V, about 8V to about 10V, about 10V to about 12V, and any and all increments therein. Embodiments of the electrical potential are applied as sinusoidal electrical waves at a frequency of about 10 MHz. The frequency may be in a range of about 0.1MHz to about 1MHz, about 1MHz to about 10MHz, about 10MHz to about 100MHz, and any and all increments therein. Voltage is applied to the DEP chip for cell capture by the positive DEP effect. An applied voltage is applied to trap a number of cells in each nanowell. The number of cells may be equal to the number of electrode pairs in each nanowell. In some embodiments, the number of cells is 1.

In some embodiments, as depicted in fig. 18, a second voltage may be applied to the DEP nanowell array (e.g., to a second pair of electrodes). In some embodiments, the second cell type is loaded with a second applied voltage (e.g., after capturing the first cell type and washing the remaining cells of the first cell type). In such embodiments, the number of cells is 2. In some applications of the invention, the number of cells is greater than 2.

Embodiments of the method include inverting the aligned array such that the array of microwells is located below the array of nanowells. That is, when single cell DEP capture is complete, the device is inverted. During inversion, electricity may continue to be applied to the DEP nanowell array to keep the trapped cells against gravity. When the electricity is interrupted, the trapped cells will be pulled by gravity into the aligned microwells below the nanowell.

In some embodiments, as depicted in fig. 17A and 17B, cells are loaded into a nanowell array slide, with the nanowell array positioned over the microwell array. That is, the nanowell array is oriented when the cells are loaded. In such embodiments, the inversion step is not necessary.

An embodiment of a method includes discontinuing application of electricity to an array of nanowells. That is, when the DEP trapping voltage is turned off, the loaded cells are transferred from the nanowell to the aligned microwell. In some embodiments, the loaded cells are transferred by gravity. However, positive pressure, vacuum, and/or vibration may also be used to facilitate movement from the nanowell to the microwell.

Embodiments of the method include loading a plurality of barcoded beads into a microwell such that a single bead occupies each cell-loaded microwell.

Embodiments of the methods include capturing RNA from a cell and retrieving the RNA-loaded beads. That is, the barcoded beads, lysis buffer and fluorinated oil are loaded sequentially to create a sealed array of cell bead pairs. The cells are lysed using one or more techniques, including for example, freeze-thaw lysis, thereby releasing the mRNA for single cell transcriptome sequencing.

Embodiments of the methods include sequencing the captured RNA. That is, mRNA on barcoded beads is captured by DNA oligomers on the bead surface, each containing a cellular barcode and a Unique Molecular Identifier (UMI). The captured mRNA is largely reverse transcribed to form single cell transcriptomes (STAMPs) attached to the microparticles. Captured RNA is amplified using suitable techniques known in the art, including, for example, PCR amplification of cDNA synthesized from the RNA, in some embodiments, purification and sequencing library preparation. In some embodiments, sequencing data for transcriptome alignments is performed to generate gene expression matrices for downstream data analysis. In certain aspects, this aspect of the invention allows for decoupling of cell capture and bead loading such that each can be performed in a poisson manner, and cells can be cycled back and forth on the DEP capture array until a capture rate of greater than 90% is achieved.

Experimental examples

The present invention is described in further detail by referring to the following experimental examples. These examples are for illustrative purposes only and are not intended to be limiting unless otherwise specified. Thus, the present invention should not be construed as limited in any way to the following examples, but rather should be construed to include any and all variations which become apparent as a result of the teachings provided herein.

Example 1:

background

High throughput single cell RNA sequencing (scra-seq) analysis has become a powerful tool for cell identification and classification in a variety of research fields, including immune responses, tumor microenvironment, and cancer heterogeneity. Using droplet-based packaging or sub-nanoliter well arrays to pair single cells and uniquely barcoded mRNA capture beads, it is feasible to profile gene expression of tens of thousands of single cells in a massively parallel manner and at relatively low cost. For example, a total of 14813 lung cells were mapped to 30 cell types to quantify changes in cell activity status in young and old mice, a transcription landscape (landscapes) of 17374 mouse bone marrow vessels, perivascular and osteoblast populations was obtained to show a previously unappreciated level of cellular heterogeneity within the bone marrow niche, and a map (atlas) containing 47016 CD45+ cells was collected from 8 primary breast cancers, revealing phenotypic expansion of immune cells within the tumor. Improvements in microfluidic technology, coupled with the maturity of the scRNA-seq clustering algorithm and statistical modeling, have changed the ability to dissect complex biological systems at the single cell level. In almost all high-throughput scRNA-seq methods, the use of molecular barcodes of uniquely barcoded beads is an enabling and indispensable step, which allows us to use a large number of RNA-seq workflows while maintaining single cell resolution.

Although a range of microfluidic-based platforms have been developed to improve reproducibility, throughput and sensitivity, there are some fundamental limitations. For droplet-based methods, a cell suspension, oligonucleotide barcoded beads and reagents are co-flowed into the system and then encapsulated in nanoliter droplets. Continuous workflow allows efficient single cell separation, however, also results in low cell-bead pairing efficiency, which prevents its adoption of low input samples. Furthermore, the inevitable requirements of peripheral devices limit the portability of remote clinical centers. Alternatively, a microtrap or nanotrap-based device provides a simplified strategy that is less costly, more portable, and potentially compatible with highly limited patient samples. By specifically adjusting the well size, single cells and barcoded beads can be co-isolated into sub-nanoliter wells with bead loading efficiencies in excess of 95%. However, to minimize the rate of two or more cells (called cell doubles or polyplexes) in the same well, the cells need to be loaded in a super-poisson fashion so that only less than 10% of the wells occupied by beads receive the cells.

In recent years, a number of techniques have been developed to improve the performance of single cell capture and bead loading. For example, to capture as many cells as possible from a limited sample, zheng et al designed a droplet-based system using GemCode beads that was capable of digitally counting 3' mrna of tens of thousands of single cells with a cell capture efficiency of about 50%. To avoid multiple beads in one droplet due to random packing, moon et al used a spiral channel based microfluidic platform to order highly concentrated beads at equal intervals prior to co-packing the cells, which resulted in deterministic packing of the beads and reduced barcoding errors. To minimize bead loss during Drop-seq implementation,

and colleagues redesigned the original infusion set to be compatible with the air pressure system and syringe pump, and designed with the accompanying accessoryChip for efficient recovery of mRNA capture beads. However, despite these technological improvements, all of the above methods still followed a stochastic passive approach to pairing single cells and mRNA capture beads. Recently, cheng et al developed Hydro-Seq, which allows contamination-free scRNA-Seq of Circulating Tumor Cells (CTC) with high cell capture efficiency (72.85 + -2.64%). However, the size-based separation configuration only works when the target cells (e.g., CTCs) are typically larger than the other unwanted cells (blood cells), which limits their use in a wide range of scra-seq applications.

Thus, there remains an unmet need to develop a high throughput method to capture thousands of single cells for scRNA-seq in a poisson protocol. Dielectrophoretic (DEP) trapping is an active trapping mechanism that has been extensively studied and may be a potential solution to this challenge. Since the pioneering use of DEP to separate live and dead yeast cells in Pohl and Hawk in 1966, this label-free method has been used to manipulate and sort bacteria and mammalian cells. Recent studies include 3D electrode-based cell separation devices that can remove 99.1% of RBCs from blood samples spiked with 1% of cancer cells at processing speeds of-170,000 cells per second; an electroactive twin-well array that can analyze intracellular material at the single-cell level with minimal loss of target cells; and a planar chip for high-throughput cell-cell pairing, with a pairing efficiency as high as 74.2%. In addition, several commercial products that incorporate DEP for cell manipulation have been used to determine particle size, isolate tumor cells, and detect biomarkers.

However, successful extension of single-cell DEP trapping to nanowell-based scRNA-seq is not trivial, including the major obstacles as follows: inherent incompatibility of optimal capture conditions between large size DNA barcoded beads and much smaller target cells. To retain sufficient surface area and barcoded oligomers to capture the entire transcriptome of a single cell, the diameter of the DNA-barcoded mRNA capture bead is typically-40 μm, and should be larger if possible. However, the DEP trapping wells must be less than 20 μm in diameter to achieve efficient and selective single-cell trapping with low probability of trapping doublets. Furthermore, the DEP trapping nanowells need to be less than 15 μm deep, since the cell trapping efficiency drops dramatically with increasing well depth. Therefore, it is not feasible to directly bind the barcoded beads and the capture of single cells in the same well for single cell RNA sequencing.

In this study, a dTNT-seq was reported, a highly integrated DEP capture nanowell transfer (dTNT) device for active capture of single cells in small nanowells 20 μm wide and transferred to larger size 50 μm wells for loading of barcoded beads to perform single cell mRNA transcriptome capture. The device consists of a top chip and a bottom chip, and cell capture and bead loading are performed in a poisson protocol, respectively. As shown schematically in fig. 1, a 50 μm microwell array slide was first pre-aligned over a 20 μm DEP nanowell array using a custom designed micro-aligner. When the single-cell DEP capture was complete, the device was inverted to transfer the captured cells to the larger microwell below after the DEP capture voltage was turned off. The striped beads, lysis buffer and fluorinated oil were then loaded sequentially to create an array of sealed cell-bead pairs and the mRNA released for single cell transcriptome sequencing using freeze-thaw lysis (see our previously reported scFTD-seq protocol). Briefly, mRNA captured onto barcoded beads was extensively reverse transcribed to form single cell transcriptomes (STAMPs) attached to microparticles, and the synthesized cDNA was then subjected to PCR amplification, purification, and sequencing library preparation. Such a configuration allows us to decouple cell capture and bead loading such that each can be performed in a poisson-like fashion, and cells can be cycled back and forth on DEP capture arrays until a capture rate of over 90% is achieved. Here, the inventors designed and fabricated an interdigitated dTNT device containing an array of 3,600 electrically active DEP traps in a single-cell trapping nanowell integrated with a matching array of large microwells. A single cell capture rate of 92% and a high transfer efficiency of 82% was demonstrated. Plus a bead loading >99%, we can break the poisson limit of cells and beads, in other words, in a double poisson distribution. Finally, the performance of the scRNA-seq was assessed by profiling a mixture of mouse fibroblast NIH3T3 cells and human embryonic kidney HEK293 cells, which was quantitatively compared to scRNA-seq data obtained using previously reported devices. dTNT seq is a successful proof of success for the dual poisson scRNA seq in high throughput (thousands of single cells per run), which is enabled by the DEP active cell capture mechanism and is due in part to the nontrivial engineering of all steps into a fully encapsulated microdevice.

Results and discussion

design of dTNT-seq device

The basic principle of DEP capture is described in the method. The dTNT device consists of an electroactive DEP nanowell array chip for single cell capture and a larger microwell chip for receiving the transferred cells and loading with DNA barcoded beads. To generate a non-uniform electric field for cell trapping, interdigitated gold electrodes with 6 μm gaps were patterned on a glass substrate. The entire surface was then coated with an SU8 insulating layer (about 10 μm thick) in which 3,600 nanowells (60 × 60) were fabricated to expose the DEP trapping electrode (fig. 2A, 2B). The diameter of the nanowell is less than 20 μm to achieve single cell trapping with negligible doublets. To confirm the quality of the microfabricated DEP trapping nanowells, a 3D optical surface profiler (NexviewTMNX 2, zygo) was used to make non-contact measurements of SU8 nanowells, showing that the nanowell structure was fully formed, matching the dimensions of the original design (fig. 2C).

The DEP nanowells dedicated to single cell capture are not large enough to accommodate the DNA barcoded beads used in the scra-seq. To address this problem, we have designed larger arrays of microwells on a polymethylmethacrylate (PMMA, also known as plexiglass or acrylic) substrate. These larger microwells are about 50 μm in diameter and depth and have a pitch of about 100 μm, which is the same size as the pitch of the DEP trap array chip. Instead of fabricating the microwells using PDMS, 50 μm microwells were fabricated directly in the SU8 layer coated on the PMMA substrate to avoid the shrinkage problem (shrinkage of about 1.5% after curing of PDMS), which may lead to difficulties in alignment with the DEP array chips. Silicon or glass may be used as the substrate material for fabricating the SU-8 microwells, although special techniques are required to drill fluid access holes through the silicon or glass. Therefore, in order to easily fabricate the inlet holes and connect the tubing for introducing cells and beads, PMMA was chosen as the substrate material for fabricating large SU-8 microwells. The two sections, DEP nanowell array chip and 50 μm microwell array chip, were precisely aligned and assembled (fig. 2D). A 100 μm thick PDMS gasket was placed between the two to form a flow channel for controlled loading of control cells, beads and reagent perfusion. Using our self-made aligner/manipulator, perfect vertical alignment of the two separate layers can be achieved to ensure efficient cell capture and transfer (fig. 2E).

Single cell capture using DEP

Mouse embryonic fibroblasts (NIH 3T 3) were used to validate the overall dNT-seq workflow and its technical performance. After first contacting (priming) the device with DEP buffer, green fluorescent stained cells (20. Mu.L, approximately 40,000 cells) were loaded through the inlet port and delivered at a flow rate of 1. Mu.L/min to fill the entire microchannel. As the cell suspension flowed across the entire nanowell array, a 10MHz 4V peak-to-peak (Vp-p) sinusoidal wave was applied to the DEP chip for cell capture by the positive DEP effect. From real-time imaging recorded by an EVOS FL automatic microscope (Life Technologies/Thermo Fisher), it was observed that DEP force was sufficient to pull cells down from the flow stream to the nanowell when the cells were near the edge of the nanowell. After 5 minutes, almost all the nanowells were gradually occupied by single cells. For low input samples (< 5,000 cells), this step can be accomplished by circulating the cells in a microfluidic chamber until all single cells are captured. The flow rate was then increased to 20 μ L/min to rinse off excess cells while keeping DEP open. We examined 35 representative regions, each containing 88 nanowells, yielding a total of 3,080 nanowells to score the number of cells in each compartment (fig. 3A). One of the captured images is shown in FIG. 3B. In a set of 3 independent experiments performed using nanowells with a depth of 10 μm, our device showed an average single cell capture rate of 91.84%, with doublets less than 2%, superior to any existing high-throughput scRNA-seq platform in single cell occupancy. Trypan blue exclusion assay after DEP capture confirmed cell viability >95%.

To evaluate the effect of nanowell depth on single cell trapping, DEP nanowells were fabricated at different depths of 5, 10, 15 and 20 μm and cell trapping experiments were performed independently. The amplitude and frequency of the applied sinusoidal AC drive voltage and the cell suspension flow rate remain the same as described above. It was found that a depth profile of 5-15 μm provided satisfactory single cell capture rates, while the probability of capturing more than one cell in a single nanowell increases with depth (fig. 3C). For a depth configuration of 20 μm, on average, more than 26% of the nanowells capture two or more cells, which is unacceptable for downstream scRNA-seq analysis. Even if the applied potential was increased to 10V (Vp-p), we observed that about 40% of the nanowells remained empty after 10 minutes of DEP trapping. The results show that DEP force decreases significantly as the depth of the nanowell increases. For 5 μm deep nanowells, little doublet was observed, as the local DEP effect precluded the accommodation of two or more cells in this dimension. However, during removal of excess cells at a flow rate of 20 μ Ι _/min, some of the captured single cells escaped from the nanowell, thereby reducing the overall single-cell capture rate. Reducing the flow rate can solve this problem, but can result in increased microfluidic capture time and reduced cell viability. However, this is still to be tested by the system. From these results, the 10 μm nanowell provides an optimal condition for single cell trapping with negligible doubling rate.

Cell transfer and bead Loading with dTNT

One key feature of the dTNT device is the use of a two-layer design for cell capture and bead loading, respectively. After DEP cell capture, the entire device was inverted (while the DEP voltage was still on, the cells were still captured on the "ceiling") to transfer the single cells into larger microwells for subsequent bead loading. Once the entire device slides, the potential is turned off. The device was left undisturbed on the bench for 10 minutes, which was sufficient to allow most of the trapped cells to leave the DEP nanowell and fall into a 50 μm microwell under the force of gravity. With this capture and transfer strategy, the dTNT device showed high transfer efficiency with an average transfer rate of 82%, despite a small number of captured cells adhering to the DEP chip after waiting 10 minutes (see fig. 3d and S1 for full images). Together with a 3,600 microwell device with a single cell capture rate of 91.84%, we transferred approximately 2,700 cells into a large microwell for scRNA-seq. The area of the DEP capture array in our current devices is small. It can be easily scaled up to capture over 10,000 single cells per run.

The rationale for designing such a modular dTNT device is that the size of the nanowell used for single cell DEP capture is too small to accommodate the barcoded beads (about 50 μm) currently widely used for scra-seq. Transfer of cells into larger microwells provides sufficient space for bead loading, subsequent co-separation with single cells, and capture of single cell-derived mRNA transcriptomes on beads. However, after inverting the device upside down, the inlet is now at the bottom of the dTNT device. To address this problem, the barcoded bead suspension was first aspirated into a tube connected to a syringe and then manually poured into the channel, rather than pipetting it onto the inlet port. Finally, excess beads were washed away and loaded with oil to prevent cross-contamination during cell lysis. Using microwell size exclusion and the ability to move beads back and forth within the flow chamber, we routinely achieved single bead loading efficiencies of >99% (see fig. 3E and figures for full images).

The dNT device outperformed other established scRNA-seq methods in terms of total bead pairing efficiency (Table 1). Specifically, in Drop-Seq, only about 5% of the bead-encapsulating droplets contain a single cell; for the InDrop method, a narrow constraint design that squeezes and slows the passage of hydrogel beads increases the pairing rate to >10%; for other nanowell-based methods, such as SeqWell, it is expected that the loaded beads will cover tens of thousands of nanowells to obtain thousands of cell-bead pairs, and therefore most beads are not used to capture single-cell derived mRNA, resulting in a large waste of expensive DNA barcoded beads. In contrast, for our dTNT device, while we still used the microwells to co-separate beads and cells, the active cell capture mechanism allowed us to break the poisson limit in the initial cell capture step and obtain >2,700 single-cell data points from as few as 3,600 bead-localized microwells (which is significantly less than required by other microwell methods, such as SeqWell). Since the cell-bead co-separation rate is about 75%, one significant advantage of our device is that it reduces bead consumption by effectively utilizing all the microwells, thereby greatly reducing cost, since barcoded beads are the most expensive reagents in current scra-seq workflow. Furthermore, although dTNT is a relatively complex integrated device, it does not necessarily increase processing time compared to existing methods. In fig. 9, the operation time of each step is listed, and the stop point is indicated. After bead loading and microwell sealing, the device can be stored at-80 ℃ for long periods of time, which allows cell/bead capture at distributed sites (e.g., small clinics or point of care), but downstream library preparation and sequencing work is done after delivery to a centralized facility.

TABLE 1 comparison of cell-bead pairing efficiency on different high-throughput scRNA-Seq platforms.

Single cell RNA profiling of mixed samples of species

To evaluate the performance of dTNT-seq for single cell transcriptome sequencing, we profiled a 1. To characterize DEP capture efficiency, HEK cells with red fluorescent dye and 3T3 cells with green fluorescent dye were loaded into the device (see fig. 4a and S4 for full images). After bead loading, mRNA capture, reverse transcription, amplification, library preparation and sequencing (nominal depth about 30,000 reads per cell), we mixed the reads with human: the mouse reference genome was aligned, yielding 2,748 single cell transcriptomes. Through a rigorous filtration process, we still obtained 2,232 high-quality single-cell transcriptomes and used this data for species discrimination analysis. In total, we obtained transcriptome profiles of 1,155 HEK cells and 1,019 3T3 cells, which are close to the expected 1. For most cells, most transcripts were aligned with the mouse or human genome, and only 2.6% (58 cells) of the identified cells had mixed phenotypes, indicating a high species-specific single-cell mRNA profile and minimal cross-contamination (fig. 4B and fig. 11A). We detected the median of 5,367 mRNA transcripts from 1,225 genes in HEK cells and 4,459 mRNA transcripts from 1,012 genes in 3T3 cells (fig. 4C, 4D).

Although DropSeq, the 10X Genomics Chemium system, and SeqWell all provided their transcriptional capture efficiency and accuracy using a mixture of HEK:3T3 species, they only sequenced hundreds of cells at the saturation depth (average 200k reads per cell), which is 5-10 times higher than our working depth. Here, to maintain consistency in cell number and sequencing depth, we compared our results with SeqWell data obtained from human Peripheral Blood Mononuclear Cells (PBMCs) and found these data to be comparable to each other (fig. 4E).

Performance comparison of non-DEP Capture methods in clustering and GSEA

Unsupervised graph-based clustering algorithms were performed to analyze our sequencing data and the results were visualized in a Unified Manifold Approximation and Projection (UMAP) graph. Following the standard pre-processing workflow of quality control, data normalization and scaling, we selected a subset of 2,000 genes that exhibited high intercellular variation in the data set to perform linear dimensionality reduction using principal component analysis (fig. 12A and 12B). Based on JackStraw plots and Elbow plots, the top 10 statistically significant PCs were used as input to project single cells onto the two-dimensional plots (fig. 12C and 12D). As shown by UMAP visualization (fig. 5 a), cells are well divided into two broad categories, each of which can be identified as a corresponding cell type as determined by species-specific genes (fig. 5B). We also generated the top 15 differentially expressed gene heatmaps for the three DEP _ human and DEP _ mouse sub-clusters, respectively (fig. 13). One problem that may arise is that our DEP capture method exposes single cells to high electric fields, which may potentially alter the transcript profile and scra-seq data. To address this concern, we compared dTNT-seq with data generated by previously reported non-DEP capture microtrap-based methods to investigate whether the use of DEP affects single cell transcriptional profiling. Single cells were randomly loaded into microwell arrays using a non-DEP-based microwell method called scFTD-seq, and then co-separated with DNA barcoded beads. In addition to the cell capture and transfer steps, both platforms used the same biochemical workflow, including freeze-thaw lysis to release mRNA, reverse transcription, PCR amplification and tagging to make sequencing libraries. The inventors performed the same experiment with a mixture of 3T3 HEK species using scFTD-seq and sequenced the library at the same depth. After filtration, 1595 single cell transcriptomes consisting of 956 mouse cells and 639 human cell types were obtained. Subsequently, UMAP cluster analysis was performed and the results showed that the main group was similar to that obtained by dTNT-seq (FIG. 5C). To explore the different gene regulatory pathways and biological processes observed for the clusters, we analyzed 4 larger subsets (human 0 and 1, mouse 0 and 1) in the scFTD seq and 6 subsets (DEP _ human 0, 1,2 and DEP _ mouse 0, 1, 2) in the dTNT seq using Gene Set Enrichment Analysis (GSEA) and compared the DEP-based and non-DEP-based methods (fig. 5D).

The GO term found to be enriched in cluster DEP _ human 0 was mainly involved in protein translation and transport (fig. 6A). The first 10 enriched gene sets include biological processes such as protein localization to the endoplasmic reticulum, protein targeting membranes, and nuclear transcription of mRNA catabolic processes. Related molecular functions of ribosomal structural components and the activity of cellular components such as ribosomes and cytoplasmic portions are also observed. GSEAs of cluster human 0 showed high overlap in top enriched GO terms when compared to each other (fig. 6B). In addition to several shared features, membrane-targeted co-translated proteins, cytoplasmic ribosomes, and translation initiation are also involved activities in the translation process. The results indicate that most cells in the DEP _ human 0 and human 0 clusters are converting the information carried in the mRNA into proteins. Furthermore, we also observed similarities between the cluster DEP _ human 1 and human 1, both controlled by transcriptional processes such as ribonucleotide binding, double stranded DNA binding and mRNA metabolic processes (fig. 14A and 14B). Here, an enrichment map of the shared GO pool is shown, and the expression pattern of the top marker gene (RPS 7) defining each cluster is also provided. The top 10 lists of GO entries of cluster DEP _ human 2 are shown in fig. 14C, which do not present a clear biological process, but metabolic events are associated with this subset of cells, as mitochondrial activity plays an essential role in cellular metabolism.

For the mouse cell cluster, DEP _ mouse 0 showed similar translation and ribosome activity as DEP _ human 0 (fig. 15A). As expected, cluster mouse 0 generated from a non-DEP device had similar enriched biological activity, with 9 of the first 10 GO entries consistent with those in DEP _ mouse 0 (fig. 15B). Similarly, the cells in DEP _ mouse 1 and mouse 1 also had nearly identical characteristics (fig. 6C, 6D). The activities significantly enriched in these two clusters are mitotic cell cycle and cell division, a process that results in the division of cellular components to form more daughter cells. Chromosomal components, i.e., cellular structures in which genes perform functions such as DNA replication, have also been identified. These observations are also confirmed by the up-regulation of the Cenpa gene, which encodes centromeric protein a, which defines the mitotic behaviour of the chromosome. In DEP _ mouse 2, the formation of extracellular structures was performed, which can not only provide the necessary physical scaffold for cellular components, but also can initiate key biochemical and biomechanical cues required for tissue morphogenesis, differentiation and homeostasis (fig. 15C).

To be relieved, these findings underscore that the use of DEP capture had little or no effect on the transcriptional profile of single cells analyzed for scRNA-seq. Thus, it demonstrates that dNT-seq is an efficient method for obtaining biologically significant data from a large single cell transcriptome without altering cell behavior and data quality. Equally important, these data indicate that even for genotypically identical samples, scRNA-seq can distinguish transcriptional states that are primarily manifested in transient cellular activities such as transcription, translation, and metabolism.

Conclusion

In order to generalize the use of scRNA-seq in biological and biomedical research and ultimately transform it for precise medical and health management, efforts are still made to further reduce costs, increase cell capture rates, increase ease of use and portability. However, current scRNA-seq technologies, such as 10x Genomics, dropSeq and SeqWell, are all based on random passive encapsulation or pairing of cells and beads, with fundamental limitations imposed by poisson statistics to prevent further improvement in cell-bead pairing efficiency. To address this challenge, dTNT-seq was developed, an active DEP capture-based method for single cell transcriptome sequencing, allowing the poisson limit to be broken. By pre-aligning the two components of the dTNT device (electrically active DEP 3,600 nanowell array and a larger size microlell array), a single cell capture rate of 91.84% was demonstrated with less than 2% doublets and an effective transfer rate of 82%. Furthermore, the compact configuration saves the consumption of expensive DNA barcoded beads by reducing the number of nanowells required to generate the same number of single-cell transcriptome data-points, which is a particular advantage of our device, compared to other microwell-based or droplet-based methods. In validation of dNT-seq, a stringent human-mouse species mixture experiment was performed. The present inventors recovered 1,155 HEK cells and 1,019 NIH3T3 cells from a single device containing as few as 3,600 nanowells. The inventors showed comparable performance in terms of the number of genes and transcripts detected per cell. Unsupervised clustering and GSEA analysis identified subtle differences in biological processes behind gene expression patterns and transcriptional status. Finally, by comparison with a non-DEP microtrap-based approach (i.e., scFTD-seq), we demonstrated that the use of DEP had little or no effect on the cell state at the transcriptional level, as evidenced by the highly consistent gene expression cluster and GO pathway shared by the clusters identified in both the dNT-seq and scFTD-seq.

DEP capture rate is a key indicator of obtaining enough single cells to perform downstream experiments and achieve high quality transcriptome profiling. To expand the application of dTNT seq, it is important to strictly assess all the operational parameters that determine the efficiency of single cell capture. Current research investigates the effect of nanowell depth and suggests that a 10 μm nanowell can provide the best single cell occupancy at a negligible double rate. Other factors, including the conductivity of the surrounding medium, the frequency and strength of the applied electric field, the diameter of the nanowell and the flow rate of the cell suspension, have been evaluated experimentally. In addition, the dielectric properties of various biological cells, such as mouse lymphocytes and erythrocytes, human erythrocytes, normal and malignant leukocytes, and leukemia cells, have also been studied. Furthermore, DEP has been successfully applied to capture and manipulate different types of cells, including single cell level human monocyte-like cells (U937), lung cancer cells (NCIH 1975), prostate cancer cells (DU 145 and PC 3), breast cancer cells (MCF-7) and HeLa cells, in addition to the 3T3 and HEK used in our studies. In conclusion, all of these contribute to the expansion of the application of dTNT-seq into a wide field of research.

Throughput is another important factor in single cell analysis. To maximize transcript identification, the goal of most scRNA-seq experiments is to analyze thousands of single cells per sample per run, taking into account the sequencing depth required for each cell to generate biologically significant data. When too many cells are pooled in a single sequencing lane, the depth of each cell is greatly reduced, and thus the number of genes detectable per cell is significantly reduced. Therefore, we designed and fabricated 3,600 nanowells for DEP trapping of single cells to verify the utility of dTNT-Seq, which is suitable for most applications. Current devices can be easily expanded to tens of thousands of chambers, if desired, to achieve massively parallel single-cell capture without significant design changes. One potential problem that may be encountered when increasing the number of nanowells per device is electrical thermal interference, including increased joule heating and local dielectric loss heating, which may affect cell viability and alter cell behavior. However, this problem can be minimized by dividing the DEP electrode array into sub-blocks while keeping the signal pads linked to the centralized control module.

The dTNT-seq method may be further modified or extended to other assays not possible with current technology. First, we can maximize the capture efficiency of precious patient samples by connecting the inlet and outlet together, circulating the cells in the flow channel. Second, analyzing cell cross-talk by sequencing physically interacting cell-cell pairs is a great challenge in tumor ecosystem research. Since cell pairs can be easily formed using DEP, we envision that this DEP-based scRNA-seq method can capture heterotypic cells by independently operating a pair of DEP traps, followed by high throughput transcriptome sequencing of cell-cell pairs, modified in a highly controlled manner to study cell-cell interactions. Examples are tumor and stromal cell interaction, tumor and macrophage interaction, and T cell and dendritic cell interaction. Overall, this DEP capture nanowell transfer strategy is an advantageous platform for profiling mRNA transcriptome from single cells or cell-cell pairs widely involved in basic or clinical biomedical research.

Method

DEP Capture theory

DEP is a phenomenon that describes the directional motion of dielectric particles in a non-uniform electric field. For spherical particles of radius r, the DEP force (F) acting on them _DEP ) Can be calculated by the following formula:

wherein ε, f and E _e Representing the absolute permittivity, activation frequency and amplitude, respectively, of the applied electric field. K is a polarization factor, which can be expressed as:

here, the number of the first and second electrodes,

and σ is the conductivity). />

And &>

The complex dielectric constants of the particles or cells and the culture medium, respectively. Re [ K (2 pi f)]Is the real part of the Clausius-Moxodi (CM) factor, which determines polarity and thus F _DEP This can be adjusted by the conductivity of the surrounding medium and the frequency of the applied electric field. For cells more polarized than the surrounding medium (i.e., re [ K (2 π f))]> 0), positive DEP attracts it towards the maximum of the electric field gradient. In cases where the cell is less polarizable (e.g., re [ K (2 π f)]< 0), negative DEP will repel it away from the high electric field gradient region. In our study, cells were captured into microwells using positive DEP forces.

Device fabrication

DEP nanowell chips were fabricated using standard soft lithography and lift-off techniques. Glass slides (Thermo Scientific;3 "x 1mm thick) were first washed with piranha solution (3 mixture of sulfuric acid and hydrogen peroxide) at 350 ℃ for 5 minutes, then rinsed with Deionized (DI) water and pre-baked. A few drops of HMDS (Microchem) were applied prior to photoresist deposition to improve adhesion. Then, a layer of AZ5214 (Microchem) capable of image reversal was spin-coated at 4000rpm for 40 seconds to form a thin film of about 1.5 μm. After soft baking at 110 ℃ for 50 seconds on a hot plate, the coating was applied at a dose of about 100mJ/cm ² Exposed to UV radiation (EVG 620 contacting/close to the mask aligner) to form a reverse pattern. The chip is then post-baked at 120 ℃ for 2 minutes and a flood exposure is applied to make the unexposed areas soluble, typically in a dose in excess of 200mJ/cm ² . Subsequently, the photoresist was developed using MF319 developer (Microchem) for 1 minute. Prior to coating of the metal layer, the chip is treated with an oxygen plasma to remove resist residues and contaminants from the surface. Then, titanium of 20nm thickness and gold of 100nm thickness were successively deposited by thermal evaporation (Kurt Lesker EJ1800 thin film deposition system). Finally, the remaining portions of the sacrificial photoresist are rinsed away with acetone along with the overlying metal. To fabricate a 10 μm deep nanowell layer on a gold electrode, SU8 2010 (Microchem) was spread onto a substrate at 500rpm for 5 seconds and then raised to 3500rpm for 45 seconds. The nanowell array pattern printed on the chrome photomask was precisely aligned with the patterned electrode and the SU8 photoresist was exposed to ultraviolet light using an MJB4 mask aligner (SUSS MicroTec) equipped with a long pass filter. After post-baking, the entire chip was developed and rinsed according to the processing guidelines and a hard bake at 150 ℃ for 5 minutes was added to the process to cure the photoresist.

For larger microwell layersFabrication, SU8 negative photoresist micromachining follows the normal workflow of the design, but the entire process is modified and optimized for a specific application using PMMA substrates (Plastics, ridout Plastics, inc.). First, PMMA of 2mm thickness was cut into a rectangle of 4.5 cm. Times.2.3 cm by a computer numerical control laser cutter. Prior to resist application, the substrate was rinsed with isopropyl alcohol (IPA) and DI water in an ultrasonic bath and then pre-baked at 60 ℃ for 3 minutes. Next, SU8 2025 (Microchem) was spin-coated at 500rpm for 5 seconds, and then at 1750rpm for 45 seconds to produce a 50 μm-thick film. The photoresist build-up on the edge of the substrate is removed by bringing a small flow of EBR PG (MicroChem) into close contact with the photomask. Soft-baking at 60 deg.C for 2 min, then soft-baking at 90 deg.C for 15 min, and exposing the substrate to 170mJ/cm ² Dose of uv light. After direct exposure, the chip was post-baked at 60 ℃ for 2 minutes, then at 90 ℃ for 15 minutes, followed by a relaxation step at 60 ℃ for 2 minutes. Development was carried out using SU8 developer (Microchem) for 15 minutes, followed by 10 seconds IPA spraying. Finally, the device inlet and outlet were drilled using a laser cutter.

Cells

NIH3T3 mouse fibroblasts were used for single cell capture and dNT-seq validation of transfer efficiency. Cells were cultured in a humidified incubator (37 ℃,5% CO) ₂ Atmosphere) in dulbecco modified Eagle medium (DMEM; gibco), glutamic acid was added to the medium and 10% fetal bovine serum (Gibco) was added. Before use, cells were detached from the bottom of the flask by applying 1mL trypsin-EDTA (Sigma Aldrich) and incubated at 37 ℃ for 3 minutes. Cells were stained with a green fluorescent probe (CellTracker Blue CMAC, invitrogen) according to the manufacturer's instructions. Briefly, 3T3 cells were plated at 2 × 10 ⁶ The cells/mL density were resuspended in serum-free DMEM containing 5. Mu.M/mL dye. After incubation in the dark at 37 ℃ for 15 minutes, the labeled cells were washed twice in complete DMEM medium. For the species mixture scRNA-seq assay, HEK cells were cultured in the same medium and stained with red fluorescent dye (CellTracker red CMTPX, invitrogen) as described above. Equal numbers of 3T3 and HEK cells were loaded prior to loadingAre mixed together.

DEP buffer

Typically, the culture medium has a high conductivity, which can induce only a negative DEP response to mammalian cells. To generate positive DEP for the capture cells, a positive DEP was prepared from 10mM HEPES, 0.1mM CaCl ₂ A low conductivity DEP buffer consisting of 59mM D-glucose and 236mM sucrose. Furthermore, 2% w/v bovine serum albumin (BSA; sigma Chemical Co.) was added to the DEP buffer to block non-specific cell adhesion. The final conductivity of the buffer was measured by a conductivity meter (EC 215, hanna Instruments) with an average reading of 272. Mu.S/cm. Notably, cell viability in this buffer has been validated by multiple reports.

Device Assembly and Experimental setup

Prior to assembly, each individual part of the dTNT device is exposed to O ₂ Plasma to make SU8 nanowell hydrophilic. A 100 μm thick PDMS gasket with a central hollow rectangular cutout was first attached to the DEP nanowell chip. The larger array of microwells was then aligned vertically above the DEP array using our own manipulator (Thorlabs, inc.) and microscope. The assembled device was secured and clamped by two PMMA plates using spring set screws. Prior to use, ethanol was slowly flowed through the device to remove air bubbles from the SU8 nanowell. Thereafter, the device filled with 5% BSA in PBS was incubated at room temperature for 30 minutes. Subsequently, the device was washed with DEP buffer prior to cell loading. During the experiment, the dTNT device was placed on the EVOS ^TM FL automated imaging system (Life Technologies) that integrates fully automated and motorized X/Y scanning stages. The system is also used to monitor and image the entire capture process. Sinusoidal potentials were induced using a function generator (SDG 1000X; siglent). A 1mL syringe (BD) connected to the outlet of the device was precisely controlled by a syringe pump (Fusion 200, chemyx inc.).

Microfluidic manipulation of dTNT seq

Prior to the experiment, 250. Mu.l (about 40,000) of the original beads were washed three times with PBS and then resuspended in 50. Mu.l of PBS; the cell suspension was centrifuged at 300g for 5 min and changed with low conductivity DEP bufferA culture medium; mu.l of a freeze-thaw lysis buffer consisting of 100mM Tris (pH 7.5), 10mM EDTA, 1M NaCl, 5. Mu.M DTT, 0.4U/mL Lucigen RNase was freshly prepared and stored on ice. The bench and microscope were carefully cleaned with rnase and 70% alcohol. Will have a density of 2X 10 ⁶ A total of 20 μ l cell suspension per cell/ml was pipetted onto the inlet reservoir and drawn into the dTNT device channel via a syringe connected to the device outlet. According to real-time imaging during DEP capture, more DEP buffer was added to remove excess cells once single cell capture reached the desired rate. The entire device was then turned upside down while keeping the DEP voltage on, after which DEP was stopped and the captured cells were allowed to fall by gravity into a larger microwell. A similar procedure for loading barcoded beads was then performed and excess beads were washed away with PBS. Next, 200. Mu.l lysis buffer was loaded and 500. Mu.l fluorinated oil (Fluorinert FC 40) was introduced into the device to seal the microwell.

Device preparation and microfluidic manipulation of scFTD-seq

Detailed materials and methods for making and operating the scFTD seq device have been described in our previous reports. Briefly, the device consists of a layer of microwell array and a layer of microfluidic channels, both made by casting PDMS on a SU8 master wafer, followed by degassing and curing at 80 ℃ for 6-8 hours. After curing, the PDMS was peeled off, the two layers cut to size, and then plasma bonded and assembled onto glass slides. Prior to cell loading, the device was pressurized using a manually operated syringe with the outlet closed to remove air bubbles within the microwell, and then first contacted with 1% BSA in PBS for 1 hour at room temperature. A total of 50. Mu.l of cell suspension (density 2X 10) ⁶ Individual cells/ml) were pipetted onto the inlet reservoir and into the device. Finally, barcoded beads, lysis buffer and fluorinated oil were loaded in sequence, similar to the case in dTNT seq.

Cytosolytic enzymes and mRNA Capture of dNT-seq and scFTD-seq

After cell and bead loading, the device was placed in a petri dish and exposed to three freeze and thaw cycles, each cycle including freezing for 10 minutes in a-80 ℃ freezer or dry ice, and then thawing for 10 minutes at room temperature. To capture mRNA onto the beads, the dTNT device was incubated in an aluminum foil covered wet chamber for 1 hour. After incubation, the device was inverted and the beads were retrieved by washing with 6 × saline-sodium citrate (SSC) buffer. Finally, the collected beads were washed twice with 6 × SSC buffer, and then subjected to a reverse transcription step.

Library preparation and sequencing

For dTNT seq and scFTD seq, library preparation was performed as described in DropSeq and SeqWell protocols (http:// mccarrollab. Com/DropSeq /). Briefly, captured mRNA was reverse transcribed using Maxima H Minus reverse transcriptase (Thermo Fisher) and custom template-switched oligonucleotides. The cDNA-coated beads were then treated with exonuclease I (Exo I, NEB) at 37 ℃ for 1 hour, spinning to chew off any unbound mRNA capture probes. The cDNA was then amplified using 13 cycle PCR whole transcriptome amplification and the cDNA library was then purified using Ampure XP beads (Beckman Coulter) at a ratio of 0.6. The sequences of the beads and all primers used in the library preparation are listed in table 2. The quality of the amplified DNA was evaluated by a bioanalyzer (Agilent inc.) using a high sensitivity chip. After standard Nextera labeling, PCR reaction (Nextera XT, illumina), another round of purification and high sensitivity bioanalyzer testing, the library was sequenced at medium depth (average 20k to 40k reads per cell) on HiSeq4000 (Illumina) and 4 samples were pooled into one sequencing lane.

TABLE 2 sequence of beads and primers used in library preparation.

Transcriptome alignment and data analysis

Raw reads were converted to digital gene expression matrices (DGE) according to the Drop-seq core calculation protocol (V2.0.0). Briefly, the 5 'adapter and 3' poly a tails were detected and trimmed to remove adapter sequences, organize the cell barcodes and UMI and match them to each gene in each cell. Pairs of reads were then aligned to the human-mouse mixed reference genome (hg 19_ mm 10) using STAR v2.5.2b. DGE was generated for cells with over 10000 readings per cell and 2748 cells were identified. The Seurat packet (V3.0) in R (V3.6) is used to perform all data analysis. For species allocation, cells with less than 500 genes, less than 2000 transcripts were considered low quality and were filtered out, and cells with greater than 90% transcript purity were considered to belong to either human or mouse species. For unsupervised clustering, the quality control criteria used to filter the cells included: 1) Gene expression counts between 200 and 5000 genes; 2) Mitochondrial gene expression was less than 10% (fig. S5b and c). After data filtering, 2,570 cells were obtained for clustering analysis of dNT-seq, leaving 1595 cells for scFTD-seq. Global scaling normalization and linear transformation ("scaling") were used as standard pre-processing steps before PCA was performed. To reduce the uncertainty in identifying the true dimensions of the dataset, the JackStraw program and the Elbow method were used to determine the top principal components to be included in the cluster.

Gene set enrichment analysis

GSEA software was used to analyze gene expression patterns for each cluster. Here, differentially expressed genes distinguishing each cluster from other clusters and corresponding fold change values were loaded into the gseaprenak. The complete Gene Ontology (GO) set, including biological processes, cellular components, and molecular functions, was used as the annotated gene set. For the Mouse cluster, the genes were transformed from the GO Gene set to the target Mouse _ Gene _ Symbol _ mapping Gene set using Chip2Chip analysis.

Example 2:

successful demonstration of dNT-seq validated that "Capture of Single cells by DEP-transfer to larger microwells-Loading barcoded beads-Capture transcriptome" could be a reliable approach to double Poisson initiative for single cells of scRNA-seq. We envision that the dTNT-seq device can be modified from three aspects to accommodate more complex and demanding applications.

Adjacent electrode pairs are designed to capture two or more types of cells to study cell-cell interactions at the single cell level

Analyzing cell cross-talk by sequencing physically interacting cell-cell pairs is also a challenge in the search of tumor ecosystems. We expect to add adjacent DEP trapping cells on the current DEP nanowell array layer so that two or more types of cells can be trapped. Figure 16 shows the design patterns and workflow of two types of cell pairing for studying cell-cell interactions by scRNA-seq. After assembling the two separate layers into an integrated device and loading the a-type cells, the first sinusoidal potential (AC 1) was turned on to capture the a-type cells into the DEP nanowell on the right. Then, B-type cells were loaded and captured into the left DEP nanowell by turning on a second sinusoidal potential (AC 2). Thereafter, the entire device can be inverted, and the captured cells transferred to the underlying larger microwell, followed by bead loading, mRNA capture, and the like. To study the interaction of more types of cells, more DEP capture units can be designed to capture each type of cell sequentially before transferring them into a larger microwell.

Design of "rooftop DEP nanowell array" devices to directly capture single cells thereon

The current dTNT-seq is a relatively complex integrated device, the complexity of which mainly stems from the "transfer" process. The design of an inverted dTNT-seq device is contemplated herein, where after initial assembly, the DEP nanowell array is at the top and the larger microwell array is at the bottom. As shown in the schematic diagram of fig. 17A, after loading type a cells into the flow channel (not shown), AC1 was opened to capture it directly on the "roof top" DEP nanowell. Then, turning off AC1 releases type a cells, allowing them to fall under gravity into the larger bottom microwell. At this point in time, the DEP nanowell can be used to capture other cells. If cell-cell interactions need to be studied, B-type cells are loaded, captured by turning on AC2, and released into a larger microtrap in a similar manner. When cell capture is complete, barcoded beads are loaded to capture mRNA for downstream sequencing and analysis. Fig. 17B shows a cross-sectional illustration of the entire process.

Addressable "rooftop DEP nanowell array" devices are designed to allow flexible manipulation of target cells

Based on the above functions, a design with control circuitry is envisioned to make each DEP nanowell programmable and addressable so that a single cell can be flexibly manipulated during the trapping process or any step that has been transferred into the bottom larger microwell. As shown in fig. 18, after capturing two types of cells, e.g., T cells (type a) and tumor cells (type B) by the "rooftop DEP nanowell array" and transferring them into the bottom larger microwell #1, cell-cell interactions can be observed or measured on a real-time imaging microscope system. If the cells in any microwell exhibit a particular state or function, e.g., T cells exhibit a high degree of cytotoxic tumor killing, we refer to them as target Cells (COIs). The exact location of the COI can be located and the corresponding DEP electric field can be turned on using a programmable control system. The selected COI can then be captured back again into the top DEP nanowell and transferred to a new larger microwell array (# 2) for further in-depth analysis.

Equivalent forms

Although preferred embodiments of the present invention have been described using specific terms, such description is for illustrative purposes only, and it is to be understood that changes and variations may be made without departing from the spirit or scope of the following claims.

Is incorporated by reference

All patents, published patent applications, and other references cited herein are expressly incorporated herein by reference in their entirety.

Sequence listing

<110> university of yale

<120> integrated dielectrophoresis-trapping and nano-trap transfer method for realizing double-poisson single-cell RNA sequencing

<130> 047162-7294WO1( )

<150> 63/027,582

<151> 2020-05-20

<160> 5

<170> PatentIn version 3.5

<210> 1

<211> 70

<212> DNA

<213> Intelligent people

<220>

<221> J _ segment

<222> (33)..(44)

<223> J-segment of barcode

<220>

<221> misc_feature

<222> (33)..(40)

<223> n is a, c, g or t

<400> 1

tttttttaag cagtggtatc aacgcagagt acnnnnnnnn tttttttttt tttttttttt 60

tttttttttt 70

<210> 2

<211> 33

<212> DNA

<213> Intelligent people

<400> 2

aagcagtggt atcaacgcag agtgaatrgr grg 33

<210> 3

<211> 23

<212> DNA

<213> Intelligent

<400> 3

aagcagtggt atcaacgcag agt 23

<210> 4

<211> 66

<212> DNA

<213> Intelligent people

<400> 4

aatgatacgg cgaccaccga gatctacacg cctgtccgcg gaagcagtgg tatcaacgca 60

gagtac 66

<210> 5

<211> 37

<212> DNA

<213> Intelligent

<400> 5

gcctgtccgc ggaagcagtg gtatcaacgc agagtac 37

Claims (11)

1. A method for single cell RNA sequencing, comprising:

aligning a micro-well array over a Dielectrophoresis (DEP) single-cell trapping nano-well array;

loading a plurality of cells into the nanowell;

applying electricity to the array of nanowells to trap a number of cells equal to the number of electrode pairs in at least one nanowell of the array of nanowells;

interrupting power to the array of nanowells to transfer the loaded cells from the nanowells to the microwells;

loading a plurality of barcoded beads into the microwell such that a single bead occupies each cell-loaded microwell;

capturing RNA from the cells and retrieving the RNA-loaded beads; and

the captured RNA was sequenced.

2. The method of claim 1, wherein the array of microtells comprises wells having a diameter of 50 μ ι η.

3. The method of claim 1, wherein the diameter of the nanowell is selected from the group consisting of: 10 μm, 15 μm and 20 μm.

4. The method of claim 1, wherein the RNA is sequenced using one or more techniques including PCR.

5. The method of claim 1, wherein the cells are loaded into the nanowell by applying a first alternating potential.

6. The method of claim 1, further comprising loading a plurality of second cell types into the nanowell.

7. The method of claim 6, wherein the second cell type is loaded using a second alternating potential.

8. The method of claim 1, further comprising:

inverting the aligned array such that the array of microtells is located below the array of nanowells.

9. A DEP trapping nano-well transfer (dTNT) system, comprising:

single cell trapping nanowell array, and

a micro-well array pre-aligned over the nano-well array, wherein the micro-well array is aligned using a micro-aligner device.

10. The dTNT system of claim 9, wherein the array of single-cell trapping nanowells comprises wells having diameters selected from the group consisting of 10 μ ι η,15 μ ι η, and 20 μ ι η.

11. The dTNT system of claim 9, wherein the micro-aligner apparatus is adapted and configured to align wells of the nano-well array with wells of the micro-well array.

Images